AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Distribution-Aware Prompt Tuning for Vision-Language Models

Eulrang Cho, Jooyeon Kim, Hyunwoo J Kim

ICCV 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Prompt learning for CLIP · 53 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add inter-dispersion loss on text prompts to CoOp+VPT (ablation)CoOp+VPT · 11 datasets, 16-shot averageaccuracy (%): 81.25 -> 81.33Improved
add intra-dispersion loss on visual prompts to CoOp+VPT (ablation)CoOp+VPT · 11 datasets, 16-shot averageaccuracy (%): 81.25 -> 81.41Improved
combine both dispersion losses (full DAPT) (ablation)CoOp+VPT · 11 datasets, 16-shot averageaccuracy (%): 81.25 -> 81.62Improved
add inter-dispersion loss alone (ablation)CoOp+VPT · Food101 16-shotaccuracy (%): 86.52 -> 86.42Hurt
add DAPT dispersion losses with 1 shot per class (ablation)CoOp+VPT · 11 datasets, 1-shot averageaccuracy (%): 61.05 -> 61.42Improved
use random-sample prototype instead of mean prototype (ablation)DAPT intra-dispersion loss · 11 datasets, 16-shot averageaccuracy (%): 81.62 -> 81.29Hurt
use random-sample prototype instead of mean prototype (ablation)DAPT intra-dispersion loss · 11 datasets, 1-shot averageaccuracy (%): 61.42 -> 65.08Improved
apply DAPT instead of CoOp for domain generalizationCLIP ViT-B/16 on ImageNet variants · ImageNet-Aaccuracy (%): 48.97 -> 48.74Hurt
apply DAPT instead of CoOp for domain generalizationCLIP ViT-B/16 on ImageNet variants · ImageNetV2accuracy (%): 64.22 -> 64.93Improved
apply DAPT on base-to-new generalization versus CoOpCLIP ViT-B/16 · OxfordPets new classesaccuracy (%): 95.43 -> 72.43Mixed
set inter-dispersion weight beta_t to 100 (ablation)DAPT · hyperparameter analysisaccuracyHurt
increase VPT learning rate from 0.002 to 0.2 (ablation)VPT · 11 datasets, 16-shot averageaccuracy (%): 68.32 -> 76.56Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy